Hugging Face's logo

oscarwu
/
mlcovid19-classifier

Text Classification
Transformers
PyTorch
TensorBoard
xlm-roberta
Generated from Trainer
Inference Endpoints
Model card Files Metrics Community
1
oscarwu commited on
Commit
1684e0b
•
1 Parent(s): ef57c35

update model card README.md

Browse files
Files changed (1) hide show
  1. README.md +35 -20
README.md CHANGED
@@ -14,15 +14,15 @@ should probably proofread and complete it, then remove this comment. -->
14
 
15
  This model is a fine-tuned version of [oscarwu/mlcovid19-classifier](https://huggingface.co/oscarwu/mlcovid19-classifier) on an unknown dataset.
16
  It achieves the following results on the evaluation set:
17
- - Loss: 0.5651
18
- - F1 Macro: 0.6566
19
- - F1 Misinformation: 0.9336
20
- - F1 Factual: 0.8316
21
- - F1 Other: 0.2048
22
- - Prec Macro: 0.6775
23
- - Prec Misinformation: 0.9344
24
- - Prec Factual: 0.7907
25
- - Prec Other: 0.3075
26
 
27
  ## Model description
28
 
@@ -41,31 +41,46 @@ More information needed
41
  ### Training hyperparameters
42
 
43
  The following hyperparameters were used during training:
44
- - learning_rate: 2e-05
45
  - train_batch_size: 32
46
  - eval_batch_size: 32
47
  - seed: 42
48
- - gradient_accumulation_steps: 8
49
- - total_train_batch_size: 256
50
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
51
  - lr_scheduler_type: linear
52
- - lr_scheduler_warmup_steps: 4367
53
- - num_epochs: 60
54
 
55
  ### Training results
56
 
57
  | Training Loss | Epoch | Step | Validation Loss | F1 Macro | F1 Misinformation | F1 Factual | F1 Other | Prec Macro | Prec Misinformation | Prec Factual | Prec Other |
58
  |:-------------:|:-----:|:----:|:---------------:|:--------:|:-----------------:|:----------:|:--------:|:----------:|:-------------------:|:------------:|:----------:|
59
- | 0.5055 | 3.67 | 500 | 0.3267 | 0.6006 | 0.9440 | 0.8517 | 0.0062 | 0.8132 | 0.9228 | 0.8502 | 0.6667 |
60
- | 0.0876 | 7.35 | 1000 | 0.3922 | 0.6636 | 0.9412 | 0.8533 | 0.1963 | 0.6975 | 0.9255 | 0.8729 | 0.2941 |
61
- | 0.0477 | 11.03 | 1500 | 0.4479 | 0.6715 | 0.9404 | 0.8562 | 0.2178 | 0.6939 | 0.9288 | 0.8695 | 0.2836 |
62
- | 0.0334 | 14.7 | 2000 | 0.5123 | 0.6622 | 0.9418 | 0.8515 | 0.1935 | 0.6996 | 0.9251 | 0.8732 | 0.3007 |
63
- | 0.0271 | 18.38 | 2500 | 0.5651 | 0.6566 | 0.9336 | 0.8316 | 0.2048 | 0.6775 | 0.9344 | 0.7907 | 0.3075 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
64
 
65
 
66
  ### Framework versions
67
 
68
  - Transformers 4.23.1
69
  - Pytorch 1.12.1+cu113
70
- - Datasets 2.5.2
71
  - Tokenizers 0.13.1
 
14
 
15
  This model is a fine-tuned version of [oscarwu/mlcovid19-classifier](https://huggingface.co/oscarwu/mlcovid19-classifier) on an unknown dataset.
16
  It achieves the following results on the evaluation set:
17
+ - Loss: 0.4565
18
+ - F1 Macro: 0.6596
19
+ - F1 Misinformation: 0.8776
20
+ - F1 Factual: 0.8823
21
+ - F1 Other: 0.2188
22
+ - Prec Macro: 0.7181
23
+ - Prec Misinformation: 0.834
24
+ - Prec Factual: 0.8952
25
+ - Prec Other: 0.4251
26
 
27
  ## Model description
28
 
 
41
  ### Training hyperparameters
42
 
43
  The following hyperparameters were used during training:
44
+ - learning_rate: 2e-06
45
  - train_batch_size: 32
46
  - eval_batch_size: 32
47
  - seed: 42
48
+ - gradient_accumulation_steps: 32
49
+ - total_train_batch_size: 1024
50
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
51
  - lr_scheduler_type: linear
52
+ - lr_scheduler_warmup_steps: 2165
53
+ - num_epochs: 50
54
 
55
  ### Training results
56
 
57
  | Training Loss | Epoch | Step | Validation Loss | F1 Macro | F1 Misinformation | F1 Factual | F1 Other | Prec Macro | Prec Misinformation | Prec Factual | Prec Other |
58
  |:-------------:|:-----:|:----:|:---------------:|:--------:|:-----------------:|:----------:|:--------:|:----------:|:-------------------:|:------------:|:----------:|
59
+ | 1.1986 | 0.94 | 16 | 1.0988 | 0.6120 | 0.8270 | 0.7623 | 0.2466 | 0.6567 | 0.7636 | 0.8578 | 0.3487 |
60
+ | 1.1885 | 1.94 | 32 | 1.0851 | 0.6163 | 0.8271 | 0.7645 | 0.2574 | 0.6588 | 0.7660 | 0.8557 | 0.3547 |
61
+ | 1.1538 | 2.94 | 48 | 1.0625 | 0.6198 | 0.8275 | 0.7683 | 0.2635 | 0.6593 | 0.7695 | 0.8521 | 0.3565 |
62
+ | 1.1386 | 3.94 | 64 | 1.0307 | 0.6235 | 0.8259 | 0.7722 | 0.2725 | 0.6564 | 0.7738 | 0.8452 | 0.3502 |
63
+ | 1.0935 | 4.94 | 80 | 0.9911 | 0.6276 | 0.8259 | 0.7797 | 0.2771 | 0.6549 | 0.7803 | 0.8392 | 0.3452 |
64
+ | 1.055 | 5.94 | 96 | 0.9445 | 0.6304 | 0.8271 | 0.7912 | 0.2730 | 0.6521 | 0.7893 | 0.8344 | 0.3327 |
65
+ | 0.9925 | 6.94 | 112 | 0.8945 | 0.6340 | 0.8270 | 0.8001 | 0.2749 | 0.6518 | 0.7976 | 0.8251 | 0.3327 |
66
+ | 0.9446 | 7.94 | 128 | 0.8448 | 0.6390 | 0.8303 | 0.8106 | 0.2760 | 0.6545 | 0.8088 | 0.8186 | 0.3360 |
67
+ | 0.8813 | 8.94 | 144 | 0.7970 | 0.6448 | 0.8355 | 0.8238 | 0.2752 | 0.6598 | 0.8185 | 0.8214 | 0.3395 |
68
+ | 0.8259 | 9.94 | 160 | 0.7475 | 0.6480 | 0.8405 | 0.8330 | 0.2704 | 0.6644 | 0.8243 | 0.8256 | 0.3434 |
69
+ | 0.7721 | 10.94 | 176 | 0.6971 | 0.6532 | 0.8483 | 0.8430 | 0.2684 | 0.6746 | 0.8281 | 0.8375 | 0.3583 |
70
+ | 0.7107 | 11.94 | 192 | 0.6542 | 0.6510 | 0.8527 | 0.8496 | 0.2507 | 0.6765 | 0.8290 | 0.8448 | 0.3557 |
71
+ | 0.6742 | 12.94 | 208 | 0.6126 | 0.6527 | 0.8554 | 0.8544 | 0.2484 | 0.6793 | 0.8298 | 0.8521 | 0.3560 |
72
+ | 0.6296 | 13.94 | 224 | 0.5735 | 0.6560 | 0.8603 | 0.8586 | 0.2491 | 0.6902 | 0.8298 | 0.8602 | 0.3804 |
73
+ | 0.5947 | 14.94 | 240 | 0.5416 | 0.6592 | 0.8641 | 0.8624 | 0.2512 | 0.6986 | 0.8299 | 0.8689 | 0.3970 |
74
+ | 0.5728 | 15.94 | 256 | 0.5164 | 0.6584 | 0.8678 | 0.8674 | 0.2402 | 0.7028 | 0.8312 | 0.8745 | 0.4026 |
75
+ | 0.5424 | 16.94 | 272 | 0.4950 | 0.6620 | 0.8711 | 0.8720 | 0.2428 | 0.7110 | 0.8315 | 0.8836 | 0.4178 |
76
+ | 0.5277 | 17.94 | 288 | 0.4798 | 0.6594 | 0.8727 | 0.8751 | 0.2305 | 0.7107 | 0.8316 | 0.8874 | 0.4130 |
77
+ | 0.5204 | 18.94 | 304 | 0.4679 | 0.6613 | 0.8749 | 0.8767 | 0.2323 | 0.7183 | 0.8335 | 0.8868 | 0.4346 |
78
+ | 0.5061 | 19.94 | 320 | 0.4565 | 0.6596 | 0.8776 | 0.8823 | 0.2188 | 0.7181 | 0.834 | 0.8952 | 0.4251 |
79
 
80
 
81
  ### Framework versions
82
 
83
  - Transformers 4.23.1
84
  - Pytorch 1.12.1+cu113
85
+ - Datasets 2.6.1
86
  - Tokenizers 0.13.1